510 posts tagged with "Kubernetes"
Container orchestration, Cluster API, and self-hosted control planes
AWS Load Balancer Controller Goes GA on Gateway API: Why Your Self-Hosted Ingress Bet Just Got Its Strongest Signal Yet
AWS Load Balancer Controller v3.0.0 made Gateway API generally available — the hyperscaler's own controller now targets typed Gateway/HTTPRoute CRDs instead of annotations. What that signals for self-hosted fleets, the four non-portable limits, and an Envoy-vs-Cilium pick table.
Your Buildpack Already Wrote the SBOM — Your Cluster Just Isn't Reading It
Cloud Native Buildpacks already emit a per-layer SBOM with every image, but most fleets never enforce it. Attaching that SBOM as a signed Sigstore attestation and verifying it with an admission controller turns scan-after-deploy into a deny decision before an unapproved image ever shares a node with other tenants.
CAPH Stopped Checking Your Server Type: What Hetzner's Hardware Churn Teaches About Pinning Machine Templates
CAPH v1.0.7 deleted its hardcoded server-type allowlist after Hetzner's October 2025 lineup rework retired types like cx22. A typo'd or retired type now passes kubectl apply and fails later at Hetzner's API — here is the four-part CI discipline that replaces the deleted guardrail.
Cloud Native Buildpacks Just Graduated: Why Your PaaS Build Step Shouldn't Belong to One Vendor
Cloud Native Buildpacks graduated in August 2026 with 535 contributors and 20+ production adopters. Graduation makes CNB the boring, multi-vendor foundation for a fleet's build step — just as Nixpacks' maintenance mode shows what single-vendor builders cost — but build speed still has to be earned with cache locality.
Freeze the Crash: How CRIU Container Checkpoints Turn a Crash-Looping Pod Into a Debuggable Core Dump
One kubelet API call freezes a failing container into a tarball of memory, file descriptors, and process state. What the January 2026 Checkpoint/Restore Working Group means, the exact checkpoint-debug-restore loop, and the four boundaries — TCP sockets, GPUs, sandboxed runtimes, pod-level restore — that still say no.
Your Controller Makes 11 API Calls, Then Goes Quiet: What the controller-runtime Cache Costs a Cluster API Fleet at Scale
A controller-runtime controller's API reads cost almost nothing after startup — one snapshot plus one watch per type. The bill comes due in memory and restart bursts instead: how the Reflector/Indexer pipeline works, the three cache mistakes everyone makes, and a 50/200/500-Machine sizing table for Cluster API fleets.
Daytona Went Closed-Source: What 'Self-Hostable' Means Now That the Control Plane Is Private
Daytona moved its core to a private codebase in June 2026, freezing the public repo at v0.190.0. A before/after accounting of what self-hosting still covers, a five-question checklist for any vendor's claim, and how E2B, Modal, Beam, Microsandbox, and the Nightona fork score.
Dragonfly Without the Database: Stop Shipping the Same Gigabytes to Every Node
A 10-node fleet pulling a 1.2 GB image downloads 12 GB per deploy; P2P distribution serves it once. We price the duplicate-byte tax across fleet sizes, unpack Dragonfly's no-database lightweight mode, and compare it with Spegel, Kraken, and a plain registry mirror.
GPU Autoscaling on Kubernetes With KEDA: Scaling on the One Metric HPA Can't See
HPA can't see GPU utilization — nvidia.com/gpu is an integer device count, so a vLLM pod at 8% CPU and a saturated GPU never scales. A worked KEDA pattern: dcgm-exporter to Prometheus to ScaledObject YAML, a minimal external scaler serving utilization plus queue depth, the scale-to-zero cost math, and where KEDA stops and DRA begins.